Add TaskSeq.concat overloads for seq, list, array, resizearray

[abelbraaksma]

Part of the push for good coverage of surface area functions, see #208 for an overview. This implements TaskSeq.concat overloads for seq, list, array and ResizeArray. Reasoning behind specific overloads is that multiple overloads with flex types is not possible: dotnet/fsharp#4943.

Each of these behave exactly like their Seq counterparts:

• raises on null source
• raises normal NullReferenceException when a nested sequence returns null

As before, the xml doc blibs have been taken from seq.fs and modified a bit for readability and applicability to TaskSeq.

The signatures are as follows:

static member concat: sources: TaskSeq<#TaskSeq<'T>> -> TaskSeq<'T> // already there
static member concat: sources: TaskSeq<'T seq> -> TaskSeq<'T>
static member concat: sources: TaskSeq<'T[]> -> TaskSeq<'T>
static member concat: sources: TaskSeq<'T list> -> TaskSeq<'T>
static member concat: sources: TaskSeq<ResizeArray<'T>> -> TaskSeq<'T>

TODO list:

• implement TaskSeq.concat for seq
• implement TaskSeq.concat for array
• implement TaskSeq.concat for list
• implement TaskSeq.concat for ResizeArray

[bartelink]

Looks good - nits and musings added after the fact as usual, sorry!

[bartelink]

"the given task sequence of task sequences" feels like a mouthful, and Combines does not really help (Flattens would be a great one word TL;DR intro to my mind if I didnt know what concat did)
If I was writing something random from scratch:

Yields a task sequence that unrolls each task sequence emanating from the source, concatenating them end to end,

[bartelink]

Also these are Combinators, so I guess they do Combine in that technical sense (which we all know is the best sense!)

Everything is always "new" for all TaskSeq operators (hence my campaign for a keyword like yield conveying the fact that stuff comes in, has treatments applied, and flows out, remaining lazy)
flattened is great

I'm not sure 'new' is a useful word; if there is a 'Combines' in the description, then there is something new happening.
But we are not actually making a thing that is 'new', it's ephemeral

Hence mq quest for a workd like 'renders' or 'yields' to convey that while we are combining things, there is not really anything 'new' except in the sense that any function that does not return its input results in a 'new' value.

Again, just musing...

[bartelink]

I guess Seq doesnt either but I'd like to see mention of NRE happening if there is a null in the input
esp for arrays.
But then LINQ SelectMany also does not document it and behaves the same way
So I guess that's case closed based on prior art
(I can recall writing test snippets in 2007 because of this not being written anywhere I was aware of)

But kudos for having the test in the first instance so it's all good I guess